Models Tab

Current

Analyze the active model's characteristics, training data, and technical details.

Convert the active model's name, precision, format, pruning state, and other properties.

Merge two models with basic options.

Alpha Ratio: Blend ratio (default 0.5 = 50% primary, 50% secondary; 0.2 = 80% primary, 20% secondary)
Overwrite model: Replace the primary model with the merged result
Save Metadata: Include metadata from both source models
Weights Clip: Limit merged weights to prevent saturation and artifacts
ReBasin: Use multiple merge permutations to retain more features from both models
Model Precision: Set the output precision (e.g., FP32, FP16)
Merge Device: Choose which device to use (CPU, GPU, etc.)
Replace VAE: Optionally use a different VAE in the merged model

Merge models using presets. Check the SDXL option for SDXL models and choose from preset alpha ratios instead of setting them manually.

Display details of all checkpoint models.

Generate a hash for each checkpoint model to identify models by content, independent of filename.

Search and download models from Hugging Face directly into SD.Next.

Download preview images and metadata for models missing them. Models with existing data are skipped.

Find models on CivitAI by type, keyword, and optional tags.

Select a model from search results, then choose the version and variant to download.

Check for updates to all checkpoint models.

Adjust LoRA strength to a preferred value.